We compared the results of self-estimates of physical activity obtained with a novel instrument (the Energy Expenditure Questionnaire, EEQ) to those obtained from questions typically asked in epidemiological investigations (reference method) in a cohort of 42,150 Swedish men and women, aged 18-94. In the EEQ, participants were asked to report total physical activity by estimating the total time during a typical day and night spent on different physical activity intensity levels from the lowest (corresponding to lying in bed, 0.9 Metabolic Energy Turnover; MET) to the highest, (exceeding the intensity of to shovelling snow by hand, i.e., >6 MET). As a comparison, they also estimated hours per week devoted to household chores, commuting and leisure time physical activities classified as; light, moderate and heavy. The average physical activity estimated with the EEQ was 1.36 MET or 32.6 METh/day or 2,341 kcal/day. In comparison, physical activity estimated with the reference method represented no more than 17% of this amount. The Spearman's correlation coefficient between the two measures was 0.26. Using EEQ, men reported significantly more physical activity than women (mean = 36.3 vs. 30.6 METh/day). Body mass index (BMI) > or = 25, education > or = 12 years, and age > or = 60 years were significantly associated with lower physical activity. Questions focusing on leisure time exercise and a few other selected activities provide estimates of activity that correlate poorly with self-reported total energy output from all physical activity and inactivity. Investigators need to be more explicit about which component of activity they investigate.